MOANA: Multi-Radar Dataset for Maritime Odometry and Autonomous Navigation Application

TL;DR

MOANA tackles the challenge of robust maritime perception by introducing the first maritime multi-radar dataset that fuses X-band and W-band radar with LiDAR and stereo cameras. It provides seven sequences across port and island regions, with ground-truth annotations and a ROS data publisher to enable benchmarking of odometry, SLAM, and object detection in diverse maritime settings. The paper details calibration procedures for cross-sensor alignment and presents benchmark results demonstrating that W-band radar enhances near-field perception while X-band supports long-range coverage, suggesting a hybrid fusion approach for robust navigation. Overall, MOANA offers a valuable resource for advancing maritime navigation research and sensor fusion, with future work aimed at richer temporal variation and more precise ground truth.

Abstract

Maritime environmental sensing requires overcoming challenges from complex conditions such as harsh weather, platform perturbations, large dynamic objects, and the requirement for long detection ranges. While cameras and LiDAR are commonly used in ground vehicle navigation, their applicability in maritime settings is limited by range constraints and hardware maintenance issues. Radar sensors, however, offer robust long-range detection capabilities and resilience to physical contamination from weather and saline conditions, making it a powerful sensor for maritime navigation. Among various radar types, X-band radar is widely employed for maritime vessel navigation, providing effective long-range detection essential for situational awareness and collision avoidance. Nevertheless, it exhibits limitations during berthing operations where near-field detection is critical. To address this shortcoming, we incorporate W-band radar, which excels in detecting nearby objects with a higher update rate. We present a comprehensive maritime sensor dataset featuring multi-range detection capabilities. This dataset integrates short-range LiDAR data, medium-range W-band radar data, and long-range X-band radar data into a unified framework. Additionally, it includes object labels for oceanic object detection usage, derived from radar and stereo camera images. The dataset comprises seven sequences collected from diverse regions with varying levels of \bl{navigation algorithm} estimation difficulty, ranging from easy to challenging, and includes common locations suitable for global localization tasks. This dataset serves as a valuable resource for advancing research in place recognition, odometry estimation, SLAM, object detection, and dynamic object elimination within maritime environments. Dataset can be found at https://sites.google.com/view/rpmmoana.

Figures (9)

• Figure 1: Overview of the MOANA dataset. To address the limitations of individual sensor performance in maritime environments, we enhanced the sensor system by integrating complementary sensors to improve both range and resolution. Data acquisition was carried out across diverse scenarios using multiple sensor types, providing a robust dataset with annotated labels for the development of learning-based algorithms.
• Figure 2: Modeling and real-world capture setups for two distinct configurations. The primary difference between the hardware setups is the orientation of the GNSS receiver: in the port sequence, the vessel’s forward direction aligns with the x-axis, whereas the island sequence is rotated by 90°. Additionally, the W-band radar is positioned on the right side in the port sequence and on the left side in the island sequence. Detailed configurations are provided in the calibration files.
• Figure 3: Composition of sensor types and example images. Radar sensors generate 360-degree scanning bird's-eye view images, while LiDAR provides 360-degree point cloud data. The camera system captures stereo images in the forward direction.
• Figure 4: Data organization and file structure for each sequence of MOANA dataset. GNSS poses are provided as text files. Extrinsic calibration parameters for the camera, LiDAR, and radars are defined relative to the base frame. Sensor data is available with distinguished file formats, such as PNG images. Additionally, labels for each frame are supplied as JSON files, each named according to the corresponding frame timestamp. For the stereo camera annotation, we denote the left camera as Cam0 and the right camera as Cam1 for JSON files.
• Figure 5: Map constructions from GNSS data-based accumulation of the Island sequences. The X-band radar (b) provides full coverage of the dataset region due to its $km$ level range and strong penetration capability; however, it exhibits relatively low resolution and accuracy. In contrast, the W-band radar (c) offers high-resolution sensing but is limited to detecting only $600m$ ranges. In the berthing area (highlighted by the red-colored square), the X-band radar fails to generate a detailed map. Conversely, the W-band radar (c) and LiDAR (d) produce a high-resolution map, capturing fine structural details.